oxc_semantic/
lib.rs

1//! Semantic analysis of a JavaScript/TypeScript program.
2//!
3//! # Example
4//! ```ignore
5#![doc = include_str!("../examples/semantic.rs")]
6//! ```
7
8use std::ops::RangeBounds;
9
10use oxc_ast::{
11    AstKind, Comment, CommentsRange, ast::IdentifierReference, comments_range,
12    has_comments_between, is_inside_comment,
13};
14#[cfg(feature = "cfg")]
15use oxc_cfg::ControlFlowGraph;
16use oxc_span::{GetSpan, SourceType, Span};
17// Re-export flags and ID types
18pub use oxc_syntax::{
19    node::{NodeFlags, NodeId},
20    reference::{Reference, ReferenceFlags, ReferenceId},
21    scope::{ScopeFlags, ScopeId},
22    symbol::{SymbolFlags, SymbolId},
23};
24
25#[cfg(feature = "cfg")]
26pub mod dot;
27
28#[cfg(feature = "linter")]
29mod ast_types_bitset;
30mod binder;
31mod builder;
32mod checker;
33mod class;
34mod diagnostics;
35mod is_global_reference;
36#[cfg(feature = "linter")]
37mod jsdoc;
38mod label;
39mod node;
40mod scoping;
41mod stats;
42mod unresolved_stack;
43
44#[cfg(feature = "linter")]
45pub use ast_types_bitset::AstTypesBitset;
46pub use builder::{SemanticBuilder, SemanticBuilderReturn};
47pub use is_global_reference::IsGlobalReference;
48#[cfg(feature = "linter")]
49pub use jsdoc::{JSDoc, JSDocFinder, JSDocTag};
50pub use node::{AstNode, AstNodes};
51pub use scoping::Scoping;
52pub use stats::Stats;
53
54use class::ClassTable;
55
56/// Semantic analysis of a JavaScript/TypeScript program.
57///
58/// [`Semantic`] contains the results of analyzing a program, including the
59/// [`Abstract Syntax Tree (AST)`], [`scoping`], and [`control flow graph (CFG)`].
60///
61/// Do not construct this struct directly; instead, use [`SemanticBuilder`].
62///
63/// [`Abstract Syntax Tree (AST)`]: crate::AstNodes
64/// [`scoping`]: crate::Scoping
65/// [`control flow graph (CFG)`]: crate::ControlFlowGraph
66#[derive(Default)]
67pub struct Semantic<'a> {
68    /// Source code of the JavaScript/TypeScript program being analyzed.
69    source_text: &'a str,
70
71    /// What kind of source code is being analyzed. Comes from the parser.
72    source_type: SourceType,
73
74    /// The Abstract Syntax Tree (AST) nodes.
75    nodes: AstNodes<'a>,
76
77    scoping: Scoping,
78
79    classes: ClassTable<'a>,
80
81    /// Parsed comments.
82    comments: &'a [Comment],
83    irregular_whitespaces: Box<[Span]>,
84
85    /// Parsed JSDoc comments.
86    #[cfg(feature = "linter")]
87    jsdoc: JSDocFinder<'a>,
88
89    unused_labels: Vec<NodeId>,
90
91    /// Control flow graph. Only present if [`Semantic`] is built with cfg
92    /// creation enabled using [`SemanticBuilder::with_cfg`].
93    #[cfg(feature = "cfg")]
94    cfg: Option<ControlFlowGraph>,
95    #[cfg(not(feature = "cfg"))]
96    #[allow(unused)]
97    cfg: (),
98}
99
100impl<'a> Semantic<'a> {
101    /// Extract [`Scoping`] from [`Semantic`].
102    pub fn into_scoping(self) -> Scoping {
103        self.scoping
104    }
105
106    /// Extract [`Scoping`] and [`AstNode`] from the [`Semantic`].
107    pub fn into_scoping_and_nodes(self) -> (Scoping, AstNodes<'a>) {
108        (self.scoping, self.nodes)
109    }
110
111    /// Source code of the JavaScript/TypeScript program being analyzed.
112    pub fn source_text(&self) -> &'a str {
113        self.source_text
114    }
115
116    /// What kind of source code is being analyzed. Comes from the parser.
117    pub fn source_type(&self) -> &SourceType {
118        &self.source_type
119    }
120
121    /// Nodes in the Abstract Syntax Tree (AST)
122    pub fn nodes(&self) -> &AstNodes<'a> {
123        &self.nodes
124    }
125
126    pub fn scoping(&self) -> &Scoping {
127        &self.scoping
128    }
129
130    pub fn scoping_mut(&mut self) -> &mut Scoping {
131        &mut self.scoping
132    }
133
134    pub fn scoping_mut_and_nodes(&mut self) -> (&mut Scoping, &AstNodes<'a>) {
135        (&mut self.scoping, &self.nodes)
136    }
137
138    pub fn classes(&self) -> &ClassTable<'_> {
139        &self.classes
140    }
141
142    pub fn set_irregular_whitespaces(&mut self, irregular_whitespaces: Box<[Span]>) {
143        self.irregular_whitespaces = irregular_whitespaces;
144    }
145
146    /// Trivias (comments) found while parsing
147    pub fn comments(&self) -> &[Comment] {
148        self.comments
149    }
150
151    pub fn comments_range<R>(&self, range: R) -> CommentsRange<'_>
152    where
153        R: RangeBounds<u32>,
154    {
155        comments_range(self.comments, range)
156    }
157
158    pub fn has_comments_between(&self, span: Span) -> bool {
159        has_comments_between(self.comments, span)
160    }
161
162    pub fn is_inside_comment(&self, pos: u32) -> bool {
163        is_inside_comment(self.comments, pos)
164    }
165
166    pub fn irregular_whitespaces(&self) -> &[Span] {
167        &self.irregular_whitespaces
168    }
169
170    /// Parsed [`JSDoc`] comments.
171    ///
172    /// Will be empty if JSDoc parsing is disabled.
173    #[cfg(feature = "linter")]
174    pub fn jsdoc(&self) -> &JSDocFinder<'a> {
175        &self.jsdoc
176    }
177
178    pub fn unused_labels(&self) -> &Vec<NodeId> {
179        &self.unused_labels
180    }
181
182    /// Control flow graph.
183    ///
184    /// Only present if [`Semantic`] is built with cfg creation enabled using
185    /// [`SemanticBuilder::with_cfg`].
186    #[cfg(feature = "cfg")]
187    pub fn cfg(&self) -> Option<&ControlFlowGraph> {
188        self.cfg.as_ref()
189    }
190
191    #[cfg(not(feature = "cfg"))]
192    pub fn cfg(&self) -> Option<&()> {
193        None
194    }
195
196    /// Get statistics about data held in `Semantic`.
197    pub fn stats(&self) -> Stats {
198        #[expect(clippy::cast_possible_truncation)]
199        Stats::new(
200            self.nodes.len() as u32,
201            self.scoping.scopes_len() as u32,
202            self.scoping.symbols_len() as u32,
203            self.scoping.references.len() as u32,
204        )
205    }
206
207    pub fn is_unresolved_reference(&self, node_id: NodeId) -> bool {
208        let reference_node = self.nodes.get_node(node_id);
209        let AstKind::IdentifierReference(id) = reference_node.kind() else {
210            return false;
211        };
212        self.scoping.root_unresolved_references().contains_key(id.name.as_str())
213    }
214
215    /// Find which scope a symbol is declared in
216    pub fn symbol_scope(&self, symbol_id: SymbolId) -> ScopeId {
217        self.scoping.symbol_scope_id(symbol_id)
218    }
219
220    /// Get all resolved references for a symbol
221    pub fn symbol_references(
222        &self,
223        symbol_id: SymbolId,
224    ) -> impl Iterator<Item = &Reference> + '_ + use<'_> {
225        self.scoping.get_resolved_references(symbol_id)
226    }
227
228    pub fn symbol_declaration(&self, symbol_id: SymbolId) -> &AstNode<'a> {
229        self.nodes.get_node(self.scoping.symbol_declaration(symbol_id))
230    }
231
232    pub fn is_reference_to_global_variable(&self, ident: &IdentifierReference) -> bool {
233        self.scoping.root_unresolved_references().contains_key(ident.name.as_str())
234    }
235
236    pub fn reference_name(&self, reference: &Reference) -> &str {
237        let node = self.nodes.get_node(reference.node_id());
238        match node.kind() {
239            AstKind::IdentifierReference(id) => id.name.as_str(),
240            _ => unreachable!(),
241        }
242    }
243
244    pub fn reference_span(&self, reference: &Reference) -> Span {
245        let node = self.nodes.get_node(reference.node_id());
246        node.kind().span()
247    }
248}
249
250#[cfg(test)]
251mod tests {
252    use oxc_allocator::Allocator;
253    use oxc_ast::{AstKind, ast::VariableDeclarationKind};
254    use oxc_span::{Atom, SourceType};
255
256    use super::*;
257
258    /// Create a [`Semantic`] from source code, assuming there are no syntax/semantic errors.
259    fn get_semantic<'s, 'a: 's>(
260        allocator: &'a Allocator,
261        source: &'s str,
262        source_type: SourceType,
263    ) -> Semantic<'s> {
264        let parse = oxc_parser::Parser::new(allocator, source, source_type).parse();
265        assert!(parse.errors.is_empty());
266        let semantic = SemanticBuilder::new().build(allocator.alloc(parse.program));
267        assert!(semantic.errors.is_empty(), "Parse error: {}", semantic.errors[0]);
268        semantic.semantic
269    }
270
271    #[test]
272    fn test_symbols() {
273        let source = "
274            let a;
275            function foo(a) {
276                return a + 1;
277            }
278            let b = a + foo(1);";
279        let allocator = Allocator::default();
280        let semantic = get_semantic(&allocator, source, SourceType::default());
281
282        let top_level_a =
283            semantic.scoping().get_binding(semantic.scoping().root_scope_id(), "a").unwrap();
284
285        let decl = semantic.symbol_declaration(top_level_a);
286        match decl.kind() {
287            AstKind::VariableDeclarator(decl) => {
288                assert_eq!(decl.kind, VariableDeclarationKind::Let);
289            }
290            kind => panic!("Expected VariableDeclarator for 'let', got {kind:?}"),
291        }
292
293        let references = semantic.symbol_references(top_level_a);
294        assert_eq!(references.count(), 1);
295    }
296
297    #[test]
298    fn test_top_level_symbols() {
299        let source = "function Fn() {}";
300        let allocator = Allocator::default();
301        let semantic = get_semantic(&allocator, source, SourceType::default());
302        let scopes = semantic.scoping();
303
304        assert!(scopes.get_binding(scopes.root_scope_id(), "Fn").is_some());
305    }
306
307    #[test]
308    fn test_is_global() {
309        let source = "
310            var a = 0;
311            function foo() {
312            a += 1;
313            }
314
315            var b = a + 2;
316        ";
317        let allocator = Allocator::default();
318        let semantic = get_semantic(&allocator, source, SourceType::default());
319        for node in semantic.nodes() {
320            if let AstKind::IdentifierReference(id) = node.kind() {
321                assert!(!semantic.is_reference_to_global_variable(id));
322            }
323        }
324    }
325
326    #[test]
327    fn type_alias_gets_reference() {
328        let source = "type A = 1; type B = A";
329        let allocator = Allocator::default();
330        let source_type: SourceType = SourceType::default().with_typescript(true);
331        let semantic = get_semantic(&allocator, source, source_type);
332        assert_eq!(semantic.scoping().references.len(), 1);
333    }
334
335    #[test]
336    fn test_reference_resolutions_simple_read_write() {
337        let alloc = Allocator::default();
338        let target_symbol_name = Atom::from("a");
339        let typescript = SourceType::ts();
340        let sources = [
341            // simple cases
342            (SourceType::default(), "let a = 1; a = 2", ReferenceFlags::write()),
343            (SourceType::default(), "let a = 1, b; b = a", ReferenceFlags::read()),
344            (SourceType::default(), "let a = 1, b; b[a]", ReferenceFlags::read()),
345            (SourceType::default(), "let a = 1, b = 1, c; c = a + b", ReferenceFlags::read()),
346            (SourceType::default(), "function a() { return }; a()", ReferenceFlags::read()),
347            (SourceType::default(), "class a {}; new a()", ReferenceFlags::read()),
348            (SourceType::default(), "let a; function foo() { return a }", ReferenceFlags::read()),
349            // pattern assignment
350            (SourceType::default(), "let a = 1, b; b = { a }", ReferenceFlags::read()),
351            (SourceType::default(), "let a, b; ({ b } = { a })", ReferenceFlags::read()),
352            (SourceType::default(), "let a, b; ({ a } = { b })", ReferenceFlags::write()),
353            (SourceType::default(), "let a, b; ([ b ] = [ a ])", ReferenceFlags::read()),
354            (SourceType::default(), "let a, b; ([ a ] = [ b ])", ReferenceFlags::write()),
355            // property access/mutation
356            (SourceType::default(), "let a = { b: 1 }; a.b = 2", ReferenceFlags::read()),
357            (SourceType::default(), "let a = { b: 1 }; a.b += 2", ReferenceFlags::read()),
358            // parens are pass-through
359            (SourceType::default(), "let a = 1, b; b = (a)", ReferenceFlags::read()),
360            (SourceType::default(), "let a = 1, b; b = ++(a)", ReferenceFlags::read_write()),
361            (SourceType::default(), "let a = 1, b; b = ++((((a))))", ReferenceFlags::read_write()),
362            (SourceType::default(), "let a = 1, b; b = ((++((a))))", ReferenceFlags::read_write()),
363            // simple binops/calls for sanity check
364            (SourceType::default(), "let a, b; a + b", ReferenceFlags::read()),
365            (SourceType::default(), "let a, b; b(a)", ReferenceFlags::read()),
366            (SourceType::default(), "let a, b; a = 5", ReferenceFlags::write()),
367            // unary op counts as write, but checking continues up tree
368            (SourceType::default(), "let a = 1, b; b = ++a", ReferenceFlags::read_write()),
369            (SourceType::default(), "let a = 1, b; b = --a", ReferenceFlags::read_write()),
370            (SourceType::default(), "let a = 1, b; b = a++", ReferenceFlags::read_write()),
371            (SourceType::default(), "let a = 1, b; b = a--", ReferenceFlags::read_write()),
372            // assignment expressions count as read-write
373            (SourceType::default(), "let a = 1, b; b = a += 5", ReferenceFlags::read_write()),
374            (SourceType::default(), "let a = 1; a += 5", ReferenceFlags::read_write()),
375            (SourceType::default(), "let a, b; b = a = 1", ReferenceFlags::write()),
376            (SourceType::default(), "let a, b; b = (a = 1)", ReferenceFlags::write()),
377            (SourceType::default(), "let a, b, c; b = c = a", ReferenceFlags::read()),
378            // sequences return last read_write in sequence
379            (SourceType::default(), "let a, b; b = (0, a++)", ReferenceFlags::read_write()),
380            // loops
381            (
382                SourceType::default(),
383                "var a, arr = [1, 2, 3]; for(a in arr) { break }",
384                ReferenceFlags::write(),
385            ),
386            (
387                SourceType::default(),
388                "var a, obj = { }; for(a of obj) { break }",
389                ReferenceFlags::write(),
390            ),
391            (SourceType::default(), "var a; for(; false; a++) { }", ReferenceFlags::read_write()),
392            (SourceType::default(), "var a = 1; while(a < 5) { break }", ReferenceFlags::read()),
393            // if statements
394            (
395                SourceType::default(),
396                "let a; if (a) { true } else { false }",
397                ReferenceFlags::read(),
398            ),
399            (
400                SourceType::default(),
401                "let a, b; if (a == b) { true } else { false }",
402                ReferenceFlags::read(),
403            ),
404            (
405                SourceType::default(),
406                "let a, b; if (b == a) { true } else { false }",
407                ReferenceFlags::read(),
408            ),
409            // identifiers not in last read_write are also considered a read (at
410            // least, or now)
411            (SourceType::default(), "let a, b; b = (a, 0)", ReferenceFlags::read()),
412            (SourceType::default(), "let a, b; b = (--a, 0)", ReferenceFlags::read_write()),
413            (
414                SourceType::default(),
415                "let a; function foo(a) { return a }; foo(a = 1)",
416                //                                        ^ write
417                ReferenceFlags::write(),
418            ),
419            // member expression
420            (SourceType::default(), "let a; a.b = 1", ReferenceFlags::read()),
421            (SourceType::default(), "let a; let b; b[a += 1] = 1", ReferenceFlags::read_write()),
422            (
423                SourceType::default(),
424                "let a; let b; let c; b[c[a = c['a']] = 'c'] = 'b'",
425                //                        ^ write
426                ReferenceFlags::write(),
427            ),
428            (
429                SourceType::default(),
430                "let a; let b; let c; a[c[b = c['a']] = 'c'] = 'b'",
431                ReferenceFlags::read(),
432            ),
433            (SourceType::default(), "console.log;let a=0;a++", ReferenceFlags::read_write()),
434            //                                           ^^^ UpdateExpression is always a read | write
435            // typescript
436            (typescript, "let a: number = 1; (a as any) = true", ReferenceFlags::write()),
437            (typescript, "let a: number = 1; a = true as any", ReferenceFlags::write()),
438            (typescript, "let a: number = 1; a = 2 as const", ReferenceFlags::write()),
439            (typescript, "let a: number = 1; a = 2 satisfies number", ReferenceFlags::write()),
440            (typescript, "let a: number; (a as any) = 1;", ReferenceFlags::write()),
441        ];
442
443        for (source_type, source, flags) in sources {
444            let semantic = get_semantic(&alloc, source, source_type);
445            let a_id =
446                semantic.scoping().get_root_binding(&target_symbol_name).unwrap_or_else(|| {
447                    panic!("no references for '{target_symbol_name}' found");
448                });
449            let a_refs: Vec<_> = semantic.symbol_references(a_id).collect();
450            let num_refs = a_refs.len();
451
452            assert!(
453                num_refs == 1,
454                "expected to find 1 reference to '{target_symbol_name}' but {num_refs} were found\n\nsource:\n{source}"
455            );
456            let ref_type = a_refs[0];
457            if flags.is_write() {
458                assert!(
459                    ref_type.is_write(),
460                    "expected reference to '{target_symbol_name}' to be write\n\nsource:\n{source}"
461                );
462            } else {
463                assert!(
464                    !ref_type.is_write(),
465                    "expected reference to '{target_symbol_name}' not to have been written to, but it is\n\nsource:\n{source}"
466                );
467            }
468            if flags.is_read() {
469                assert!(
470                    ref_type.is_read(),
471                    "expected reference to '{target_symbol_name}' to be read\n\nsource:\n{source}"
472                );
473            } else {
474                assert!(
475                    !ref_type.is_read(),
476                    "expected reference to '{target_symbol_name}' not to be read, but it is\n\nsource:\n{source}"
477                );
478            }
479        }
480    }
481}
oxc_semantic/lib.rs

oxc_semantic/
lib.rs
